Claude Fable 5

Anthropic's smartest model — Mythos class

What it's the best tool for

Smartest Claude ever shipped: Mythos class, above Opus
Finishes long multi-step tasks on its own, no hand-holding
1M-token context — about 1,500 pages of documents at once
Verifies and fixes its own mistakes before answering
Reads files and images: contracts, reports, scans, slide decks
Record scores on real-work tests: 80.3% on SWE-Bench Pro

Limitations

When to reach for something else

Flagship pricing — costs more than other Claude models
Slower than lightweight models: it thinks before answering
No image or video generation — text output only
Strict Anthropic safety filters on sensitive topics

Sample output

How Claude Fable 5 responds

Prompt

Here's a year of customer reviews and our price list. Work out what clients praise and complain about, then build a plan: what to change in our services and pricing to grow revenue.

Claude Fable 5

Analysed 2,400 reviews. Key signal: speed gets praised (38% of mentions), manager availability gets criticised (22%). 1. Launch a Priority tier with a dedicated manager — fixes the top complaint and adds 12–18% to average order value. 2. Retire the entry tier: it drives 60% of complaints but only 9% of revenue. 3. Raise the base tier by 10% — pricing is almost never criticised in reviews. Below: a 90-day rollout plan with metrics to track.

Where teams use it

Four scenarios where it pays for itself

End-to-end business tasks

Reports, strategies, documents — like a chief of staff

Document work

Reads 1,000+ pages of contracts and surfaces every risk

Marketing and sales

Research, copy and competitor breakdowns in minutes

Long projects

Runs multi-step work on its own and verifies the outcome

About model

More about Claude Fable 5

Claude Fable 5 — the Anthropic AI That Works Like a Whole Team

Claude Fable 5 is the first model of the new Mythos class — a tier above Claude Opus and the smartest AI Anthropic has ever made generally available. Earlier models were assistants you had to guide step by step; Fable 5 behaves more like a strong hire: you describe the outcome, and the model plans the work, executes it, double-checks itself and delivers a finished result.

What it means in practice

Fable 5 takes on tasks that cost a person hours or days and carries them through without constant prompting. Upload an annual report, a stack of contracts or a CRM export — the model holds up to 1 million tokens in memory, roughly 1,500 pages of text. It reads documents, spreadsheets and images, so scans, screenshots and slide decks are fair game too.

A real-world example: at Stripe, the model compressed five months of engineering work into days. The same principle applies to everyday business tasks — a marketing plan, a contract review, a financial model: assign the task in the evening, collect the result in the morning.

How Fable 5 differs from other Claude models

On benchmarks built around real work, Fable 5 clearly outperforms both Claude Opus and competitors: 80.3% on SWE-Bench Pro versus 69.2% for Opus 4.8 and 58.6% for GPT-5.5. For you that means fewer mistakes and fewer redos — the model verifies its own conclusions before handing them over and fixes the issues it finds.

Who it's for

Founders — report analysis and growth strategy. Lawyers — contract review and risk hunting across hundreds of pages. Marketers — market research, content plans, copy. Finance teams — models and forecasts. Fable 5 understands plain language: brief it the way you would brief an employee.

Try it on NetRoom

NetRoom gives you Claude Fable 5 online, right in the browser: no separate $20 subscription, one balance for every top model in the catalog, pay only for what you actually use. Draft with a cheaper model, then hand the final pass to Fable 5. Start with one task you keep postponing — and see how fast Fable 5 gets it done.